You can take that one step further: all stats about the Coronavirus case counts are worthless. There is no country that has done a large enough random antibody test to be able to make a definitive statement about how far the virus has gone in infecting the population and the speed with which the situation develops is so high that any published numbers are outdated by the time they are committed to print or bits.

I don't agree with that. There's a spectrum between information being entirely uninformative and perfectly informative. Daily case headlines are pretty close to uninformative on their own, but the underlying case figures, in bulk, plugged into a model? It's certainly better than sticking your finger in the air.

Predictions don't have to be perfect to be useful, and while there is a debate to be had where the threshold is between a prediction that's confident to be useful and one that isn't, writing _all_ predictions and _all_ stats off because they're not perfect is a bit extreme.

That would require uniform reporting requirements between different countries, would require the healthcare system not to be so overloaded that they are not going to be able to comply with their reporting duties and so on. Sticking your finger in the air may actually be the preferred method.

A model's outputs can only ever be as good as its inputs. If you want to base your policy on what is effectively random noise then you are going to be wrong no matter what.